94
R.O. Dubayah, E.F. Wood, E.T. Engman et al.
5.4.3 Precipitation
Estimation of precipitation by satellite remote sensing remains problematic at the
time scales required by hydrologic models. While missions such as TRMM have
great promise for improving climatological estimates of precipitation, and perhaps
for estimating areal precipitation at seasonal to annual time steps, the current (and
planned) platforms suffer from an inability to observe the diurnal cycle directly. In
the U.S., the new NOAA WSR-88D weather radars are now producing archived
precipitation products at 4 Ian resolution. The most widely available (level 3)
products effectively incorporate some observing station data as well. Arola et al.
(1994) have used these products with some success over part of the Arkansas-Red
basin. Known problems with these data products, such as a radial bias in the estimated intensities, are currently being addressed (e.g., Smith, 1996). In the central
U.S. and other areas without complex topography, it seems likely that such products will eventually become a standard forcing for hydrologic models. In mountainous areas, such as the western U.S., radar-based precipitation products do not
appear to be viable for macro scale modeling. More detailed information on the
potential of utilizing remote sensing for precipitation estimation is given in Chaps.
6, 16 and 18.
5.4.4 Air Temperature
An innovative approach has been developed that estimates near surface air temperature from A VHRR data. The TemperatureNegetation Index (TVX) utilizes
the relationship between the normalized difference vegetation index (NDVI) and
surface skin temperature (Goward et al. 1994, Prince and Goward 1995, Prihodko
and Goward 1997, Czajkowski et al. 1997, Prince et al. 1998). The surface temperature sensed by A VHRR is some function of canopy and soil background contributions. A 9 Ian x 9 Ian (or other sufficiently sized) window centered on the
pixel for which an air temperature is to be estimated is used to generate a regression between NDVI and surface temperature and extended to a maximum NDVI.
The window is moved to cover each pixel of the image. Due to the low thermal
capacity of leaves which prevents them from reaching temperatures much higher
than about 2°C from air temperature (Gates 1980), the temperature of the canopy
at maximum NDVI will approximately be air temperature. An underlying assumption is that air temperature does not vary as rapidly in space across the spatial
array. For areas where the TVX relationship is contaminated by clouds or the
variability of NDVI in the landscape is low, such as during winter, no air temperature may be inferred.
There are several limitations to using A VHRR data to derive air temperature.
The daily curve of air temperature cannot be estimated from A VHRR observations
directly as these are only available at overpass time during mid-afternoon. One
approach is to use the overpass temperature to scale an "average" diurnal curve
obtained from climatology (Fig. 5.3). The actual form of the curve will differ from
climatology whenever factors affecting air temperature change within a day, e.g. a
precipitation event, passage of a cold or warm air mass etc. Similarly, cloudy conditions present another limitation because surface temperature cannot be sensed,
and therefore no TVX relationship may be formed. One approach is to interpolate
R.O. Dubayah, E.F. Wood, E.T. Engman et al.
5.4.3 Precipitation
Estimation of precipitation by satellite remote sensing remains problematic at the
time scales required by hydrologic models. While missions such as TRMM have
great promise for improving climatological estimates of precipitation, and perhaps
for estimating areal precipitation at seasonal to annual time steps, the current (and
planned) platforms suffer from an inability to observe the diurnal cycle directly. In
the U.S., the new NOAA WSR-88D weather radars are now producing archived
precipitation products at 4 Ian resolution. The most widely available (level 3)
products effectively incorporate some observing station data as well. Arola et al.
(1994) have used these products with some success over part of the Arkansas-Red
basin. Known problems with these data products, such as a radial bias in the estimated intensities, are currently being addressed (e.g., Smith, 1996). In the central
U.S. and other areas without complex topography, it seems likely that such products will eventually become a standard forcing for hydrologic models. In mountainous areas, such as the western U.S., radar-based precipitation products do not
appear to be viable for macro scale modeling. More detailed information on the
potential of utilizing remote sensing for precipitation estimation is given in Chaps.
6, 16 and 18.
5.4.4 Air Temperature
An innovative approach has been developed that estimates near surface air temperature from A VHRR data. The TemperatureNegetation Index (TVX) utilizes
the relationship between the normalized difference vegetation index (NDVI) and
surface skin temperature (Goward et al. 1994, Prince and Goward 1995, Prihodko
and Goward 1997, Czajkowski et al. 1997, Prince et al. 1998). The surface temperature sensed by A VHRR is some function of canopy and soil background contributions. A 9 Ian x 9 Ian (or other sufficiently sized) window centered on the
pixel for which an air temperature is to be estimated is used to generate a regression between NDVI and surface temperature and extended to a maximum NDVI.
The window is moved to cover each pixel of the image. Due to the low thermal
capacity of leaves which prevents them from reaching temperatures much higher
than about 2°C from air temperature (Gates 1980), the temperature of the canopy
at maximum NDVI will approximately be air temperature. An underlying assumption is that air temperature does not vary as rapidly in space across the spatial
array. For areas where the TVX relationship is contaminated by clouds or the
variability of NDVI in the landscape is low, such as during winter, no air temperature may be inferred.
There are several limitations to using A VHRR data to derive air temperature.
The daily curve of air temperature cannot be estimated from A VHRR observations
directly as these are only available at overpass time during mid-afternoon. One
approach is to use the overpass temperature to scale an "average" diurnal curve
obtained from climatology (Fig. 5.3). The actual form of the curve will differ from
climatology whenever factors affecting air temperature change within a day, e.g. a
precipitation event, passage of a cold or warm air mass etc. Similarly, cloudy conditions present another limitation because surface temperature cannot be sensed,
and therefore no TVX relationship may be formed. One approach is to interpolate